.\llama_bin\llama-cli.exe -m medgemma-night-sentinel-Q4_K_M.gguf -c 2048 -n 256 -p "<start_of_turn>user`n[NIGHT SENTINEL SYSTEM]`nAnalyze the following continuous cardiac monitoring data:`nPatient baseline HR is 70 bpm. Current HR is 135 bpm with irregular rhythm. SpO2 is 92%. What is your assessment?`n<end_of_turn>`n<start_of_turn>model`n"

How to Test (Python)

python

from huggingface_hub import hf_hub_download
from llama_cpp import Llama

# Download the model from this repo
model_path = hf_hub_download(
    repo_id="Ismailea04/medgemma-night-sentinel",
    filename="medgemma-night-sentinel-Q4_K_M.gguf",
)

# Load GGUF model
llm = Llama(model_path=model_path, n_ctx=2048, verbose=False)

prompt = (
    "<start_of_turn>user\n"
    "[NIGHT SENTINEL SYSTEM]\n"
    "Patient Profile: 65yo male, post-operative recovery. Baseline HR 70.\n"
    "Event/Anomaly: Sudden HR spike to 135 bpm with irregular rhythm, SpO2 92%.\n"
    "TASK:\n"
    "1. Compare the current data to the patient's baseline (name it: ##Comparaison).\n"
    "2. Identify clinical anomalies (name it: ##Detection).\n"
    "3. Provide a short interpretation (name it: ##Interpretation).\n"
    "<end_of_turn>\n"
    "<start_of_turn>model\n"
)

response = llm(prompt, max_tokens=256, stop=["<end_of_turn>"])
print(response["choices"][0]["text"].strip())

Example Output

##Comparaison
- Baseline HR 70 bpm vs current 135 bpm (marked tachycardia).

##Detection
- Irregular rhythm suggests possible atrial fibrillation or ectopy.
- SpO2 92% indicates mild desaturation.

##Interpretation
Acute tachyarrhythmia with mild hypoxemia. Recommend prompt clinical assessment,
repeat vitals, and ECG evaluation.

Intended Use

This model is intended for research and prototyping of cardiology monitoring assistants. It is not a medical device and must not be used for clinical decision-making without qualified medical oversight.

Acknowledgements

Base model: google/medgemma-4b-it
Quantization tooling: llama.cpp